`

Jeff Berlin - Virtuoso Bassist and Educator

“I feel a large part of music education is a disaster today. A lot of it is coming from people who don’t know how to teach”

Easy

## Wild Claims Continued ?

Quickie Route to Data Scientist

• R for Business Intelligence
• Data Science for Epidemiologists
• Data Science for Nursing Research
• Intro to Statistics using R
• Data Visualization for Biologists
• Statistics for Clinicians
• Machine Learning for City Planners
• R for SAS Programmers

• Data Scientist
• Data Analyst
• Data Engineer
• Data Architect
• Chief Data Officer

• BIOS 545 Introduction to R Programming
  • For BIOS students only (for the most part)
  • Ground up introduction to R
  • Fast-paced
  • Assumes some previous programming

• BIOS 534 Introduction to Machine Learning
  • Drafted at the last minute to help out
  • Blend of theoretical and practical
  • Students don’t like the theoretical yet it’s important
  • The practical draws together lots if approaches
  • Everyone thinks Data Science is all Deep Learning

• Nursing 741 Big Data Analytics
  • Analyze ethical issues related to the use of big data in health related research
  • Understand the principles of reproducible research
  • Analyze ethical issues related to big data analytics in health related research

• LAW 890 Fundamentals of Innovation (https://tigerinnovation.org/)

TI:GER, brings together graduate students in law, business, science, and engineering to work on start-up projects to transform highly promising research into economically viable projects.

Most of what I do (or try to) is based on the Lean Start-up Concept which relies heavily on “customer discovery”.

• Seventy five percent of ALL business start-ups fail
• Cool technology and ideas are frequently NOT ready for market
• Large Business Plans are conceived using stale market research
• No one “really” wants to talk to customers

Lean Start-up avoids this by:

• Frequent hypothesis testing over elaborate planning
• Customer feedback over intuition or “we know what’s best”
• Iterative Design (e.g. Fail towards success) over Big Design
• Frequent “pivots” into new directions

https://leanlaunchpad.stanford.edu/ https://steveblank.com/slides/lean-startup/

Using this approach I have:

• Obtained a patent
• Gotten funding from Georgia Research Alliance
• Active Intellectual Property declarations with Emory OTT
• Function as Entrepreneurial Lead for a Natural Language Processing recruitment tool

• The course is required
• It will be useful for later work
• Someone advised me to take this class
• I couldn’t find a more interesting course to take

• Less reliance upon others
• Enhanced productivity
• Eliminate barriers to research
• Improve employment prospects

You Are Really Facilitating Independence

• As a teacher or student you SHOULD have this goal
• You Need to know the material
• Better Yet, you need to know how to explain it

In step 2, replace “sixth grader” with a term corresponding to a beginner relative to the topic of interest (e.g. “first year”)

“STOP PLAYING ALL THOSE WEIRD NOTES, PLAY THE MELODY”

• Focus on the basics and build into the “edge” cases.
• Don’t start with oblique or uncommon topics
• Explain exceptions AFTER having laid the foundation

“JUST BECAUSE YOU’RE NOT A DRUMMER, DOESN’T MEAN YOU DON’T HAVE TO KEEP TIME”

• Everyone is responsible for maintaining a course rhythm

• “Keeping time” means establishing a pace and reliable pulse

“STAY IN SHAPE! SOMETIMES A MUSICIAN WAITS FOR A GIG, & WHEN IT COMES, HE’S OUT OF SHAPE & CAN’T MAKE IT”

Being “in shape” refers to readiness for playing (not necessarily physical shape). Do your rehearsals and practice so you are ready for action.

In a linear regression model, how does \(R^2\) differ from Adjusted \(R^2\) ?

\[ RSS = \sum_{i=1}^n (y_i - f(x_i))^2 \] \[ TSS=\sum_{i=1}^{n}\left(y_{i}-\bar{y}\right)^2 \] \[ R^2 = 1 - {RSS\over TSS} \ \] \[ R^2 Adjusted = 1 - {{(1 - R^2)(N - 1)}\over N - p - 1} \] Discussion ? Where to Start ? Recitation of Learned Facts ?

• R,SAS, SPSS are analysis frameworks first and programming languages second
• Python and Java are programming languages first with analysis ad-dons
• MATLAB is a hybrid

This doesn’t include more specific tools

• Weka
• TensorFlow
• KNIME
• RapidMiner
• Orange
• Pandas

• Apache Spark
• Snowflake
• Data Bricks
• TensorFlow
• Big Table
• Redshift
• Apache Mahout

Many of these have APIs (Application Programming Interfaces) that let you use your programming language of choice as a primary Interface

• Sweave
• knitr
• Markdown
• Jupyter

• Great for Backing Up Your Code
• Can handle multiple versions
• Can Share with others
• Others can modify and submit changes
• Issue Tracking
• Contains a Wiki

Like Learning to Cook

• Start with a good recipe
• Demonstration (by a competent cook)
• Observation (by students)
• Duplication (basic repetition)
• Imitation (repetition with changes)
• Fear of Messing Things Up

• Unhealthy
• Little Variation
• Expensive
• No Self Sufficiency
• No Independence

• For advanced students give them the project all at once

• For Intro classes cover each subtopic per week + Homework and Labs + Maybe a final project

• For Intermediate classes do a combination + Have mid term and final projects + Some supporting homework

Pasta

Solving for \(\sqrt{S}\) is the same as solving \(f(x) = x^2 - S = 0\)

We can use Newton’s Method to iterate towards an answer:

\[x_{n+1} = x_{n} - \frac{f(x)_n}{f'(x)_n} = x_n - \frac{x_{n}^2 - S}{2x_n} = \frac{1}{2}(x_n + \frac{S}{x_n})\] Make a first guess and then compute \(x_{n+1}\) until \(x_{n+1}^2\) is close enough to S within some specified tolerance

Make a guess (e.g 2) of the square root of some number (e.g. 16)
Specify a tolerance level (e.g. 0.0001)

while the absolute value of the square of that guess minus 
      the number  is >= than tolerance
          compute a new guess

somenum <- 16
tolerance <- 0.0001
guess <- 2

# Note we have to use Absolute value since we 
# are concerned with the magnitude of the difference

while( abs((guess^2)-somenum) >= tolerance) {
  guess <- (guess + (somenum/guess)) * 0.5
}
guess

somenum <- 16
tolerance <- 0.0001
guess <- 2

# Write a Function to judge quality of computed guess 

compare <- function(guess,target) {
   diff <- abs((guess^2)-target)  
   return(diff)
}

while(compare(guess,target) >= tolerance) {
  guess <- (guess + (target/guess)) * 0.5
}

guess

mySqrt <- function(target=16,guess=2,tolerance=0.0001,verbose=FALSE) {
  while( abs((guess^2)-target) >= tolerance) {
    guess <- (guess + (target/guess)) * 0.5
    if (verbose) {
      print(guess)
    }
  }
  return(guess)
}

mySqrt <- function(target=16,guess=2,tolerance=0.0001,verbose=FALSE) {
  compare <- function(guess,target) {
   diff <- abs((guess^2)-target)  
   return(diff)
  }
  
  while(compare(guess,target) >= tolerance) {
    guess <- (guess + (target/guess)) * 0.5
    if (verbose) {
      print(guess)
    }
  }
  return(guess)
}
mySqrt(16,2,0.0001)

sapply(c(16,22,49,39),mySqrt)

# or

round(sapply(c(16,22,49,39),mySqrt),2)

# or

Sqrt <- function(input=c(16,22,49,39),rnd=2) {
  retvec <- round(sapply(input,mySqrt),rnd)
  return(retvec)
}

• Lectures on Honor Code Are Helpful
• Keeps the Honest Person Honest
• It Can Still Happen
• Google Is Easy To Use
• Stack Overflow Knowingly Aids and Abets Cheating
• Cheating is Easy to Spot

• Competition
• Start the assignment late
• Later assignments rely on earlier work
• Very easy to get help from others
• Students circulate past assignments
• Claims of language confusion

In 2011 this assignment was given in the "Paradigms for Computing’’ class at Stanford

Within 1 day this was found on the Stack Overflow site: